[GRASSLIST:3212] Re: GRASS-Python, or, Alber'salgorithm?/r.proj with big files

[Len Coop]

Patrick,

I forget the number of rows-cols that I was able to re-project when I bumped my
PIII 550 box
from 256 MB to 384 MB, but it was less than the 20000x24000 (90 MB) file
I would like to reproject at the moment. 

My new box (PIV 1.9 Ghz) has .5 GB Rambus memory,
which would be expensive to toss and replace with 1 GB, but I suspect that
would NOT be enough for these NLCD files. As it goes now, I CRASH the box
running r.proj, requiring a cold boot, which I have never run into before
with Linux (the process should die and not take the whole operating system
down with it). Here is the final thing I see (this after downsizing output 
resolution) before the box freezes:

GRASS:~ > r.proj input=NLCD.OR location=nw_cd
Input:
Cols:   23924 (23924)
Rows:   20221 (20221)
North: 2907510.000000 (2907510.000000)
South: 2300940.000000 (2300940.000000)
West:  -2301360.000000 (-2301360.000000)
East:  -1583700.000000 (-1583700.000000)
ew-res: 29.997492
ns-res  29.997033
 
Output:
Cols:   12358 (12391)
Rows:   10508 (10508)
North: 46.989583 (46.989583)
South: 41.735417 (41.735417)
West:  -125.065565 (-125.065565)
East:  -115.453673 (-115.428007)
ew-res: 0.000778
ns-res  0.000500
Allocating memory and reading input map...   66%


So I maintain that r.proj seems to have big problems with big files.

Len

Patrick Dunlavey wrote:
> 
> At 10:07 AM 2/25/2002 -0800, Len Coop wrote:
> >Anybody care to re-write r.proj so it uses memory more efficiently? That has
> >been my #1 gripe about GRASS these days.
> 
> I haven't had problems reprojecting sizable datasets (e.g. 4800x1600 cells
> latlong to 4000x3000 cells albers equal area) on a RedHat 7.0 box with 128
> mb RAM and PentiumPro 200mhz. It's slow, and this is not the machine I
> would work on by choice, but it doesn't choke.
> 
> Two thoughts: If r.proj is choking due to a lack of memory, with the price
> of memory chips these days, I don't think it should be hard to justify
> bumping up your RAM.
> 
> Alternatively, I wonder if there are some factors that influence success
> with r.proj besides available memory (e.g. could be that different source
> and destination projections require different resources).
> 
> -- Pat

-- 
Leonard Coop
Research Associate
Entomology Dept & Integrated Plant Protection Center
2046 Cordley
Oregon State University
Corvallis OR 97331-2907
Phone 541-737-5523
Fax   541-737-3080
http://osu.orst.edu/dept/entomology/coopl